Insulin is a potent metabolic and growth promoting hormone that acts on cells to stimulate glucose, protein, and lipid metabolism, as well as RNA and DNA synthesis. A well-known effect of insulin is the regulation of glucose levels in the body. This effect occurs predominantly in liver, fat, and muscle tissue. In the liver, insulin stimulates glucose incorporation into glycogen and inhibits the production of glucose. In muscle and fat tissue, insulin stimulates glucose uptake, storage, and metabolism. Defects in glucose utilization are very common in the population, giving rise to diabetes.
Insulin initiates signal transduction in target cells by binding to a specific cell-surface receptor, the insulin receptor (IR). The binding leads to conformational changes in the extracellular domain of IR, which are transmitted across the cell membrane and result in activation of the receptor's tyrosine kinase activity. This, in turn, leads to autophosphorylation of tyrosine kinase of IR, and the binding of soluble effector molecules that contain SH2 domains such as phophoinositol-3-kinase, Ras GTPase-activating protein, and phospholipase Cy to IR (Lee and Pilch, 1994, Am. J. Physiol. 266:C319-C334).
Insulin-like growth factor 1 (IGF-1) is a small, single-chain protein (MW=7,500 Da) that is involved in many aspects of tissue growth and repair. Recently, IGF-1 has been implicated in various forms cancer including prostate, breast, colorectal, and ovarian cancers. It is similar in size, sequence, and structure to insulin, but has 100-1,000-fold lower affinity for IR (Mynarcik et al., 1997, J. Biol. Chem. 272:18650-18655).
Clinically, recombinant human IGF-1 has been investigated for the treatment of several diseases, including type I diabetes (Carroll et at., 1997, Diabetes 46:1453-1458; Crowne et al., 1998, Metabolism 47:31-38), amyotropic lateral sclerosis (La! et al, 1997, Neurology 49:1621-1630), and diabetic motor neuropathy (Apfel and Kessler, 1996, CIBA Found. Symp. 196:98-108). Other potential therapeutic applications of IGF-1, such as osteoporosis (Canalis, 1997, Bone 21:215-216), immune modulation (Clark, 1997, Endocr. Rev. 18:157-179) and nephrotic syndrome (Feld and Hirshberg, 1996, Pediatr. Nephrol. 10:355-358), are also under investigation.
A number of studies have analyzed the role of endogenous IGF-1 in various disease states. Interestingly, several reports have shown that IGF-1 promotes the growth of normal and cancerous prostate cells both in vitro and in vivo (Angelloz-Nicoud and Binoux, 1995, Endocdnol. 136:5485-5492; Figueroa et at., 1995, J. Clin. Endocfinol. Metab. 80:3476-3482; Torring et al., 1997, J. Urol. 158:222-227). Additionally, elevated serum IGF-1 levels correlate with increased risks of prostate cancer, and may be an earlier predictor of cancer than is prostate-specific antigen (PSA) (Chan et al., 1998, Science 279:563-566). Recent studies have indicated a connection between IGF-1 levels and other cancers such as breast, colorectal, and ovarian. Serum IGF-1 levels are regulated by the presence of IGF binding proteins (IGFBP) which bind to IGF-1 and prevent its interaction with the IGF-1 receptor (IGF-1R; reviewed in Conover, 1996, Endocr. J. 43S:S43-S48 and Rajaram etal., 1997, Endocr. Rev. 18:801831). Interestingly, PSA has been shown to be a protease that cleaves IGFBP-3, resulting in an increase of free IGF-1 in serum (Cohen et al., 1992, J. Clin. Endocrinol. Metab. 75:1046-1053; Cohen et a!., 1994, J. Endocrinol. 142:407415; Lilja, 1995, J. Clin. Lab. Invest. Suppl. 220:47-56). Clearly, regulation of IGF-1R activity can play an important role in several disease states, indicating that there are potential clinical applications for both IGF-1 agonists and antagonists.
IGF-1R and IR are related members of the tyrosine-kinase receptor superfamily of growth factor receptors. Both types of receptors are composed of two α and two β subunits which form a disulfide-linked heterotetramer (β-α-α-β). These receptors have an extracellular ligand binding domain, a single transmembrane domain, and a cytoplasmic domain displaying the tyrosine kinase activity. The extracellular domain is composed of the entire a subunits and a portion of the N-terminus of the β subunits, while the intracellular portion of the β subunits contains the tyrosine kinase domain. Another family member is insulin-related receptor (IRR), for which no natural ligand is known.
While similar in structure, IGF-1R and IR serve different physiological functions. IR is primarily involved in metabolic functions whereas IGF-1R mediates growth and differentiation. However, both insulin and IGF-1 can induce both mitogenic and metabolic effects. Whether each ligand elicits both activities via its own receptor, or whether insulin exerts its mitogenic effects through its weak affinity binding to IGF-1R, and IGF-1 its metabolic effects through IR, remains controversial (De Meyts, 1994, Horm. Res. 42:152-169).
IR is a glycoprotein having molecular weight of 350-400 kDa (depending of the level of glycosylation). It is synthesized as a single polypeptide chain and proteolytically cleaved to yield a disulfide-linked monomer α-β insulin receptor. Two α-β monomers are linked by disulfide bonds between the α-subunits to form a dimeric form of the receptor (β-α-α-β-type configuration). The a subunit is comprised of 723 amino acids, and it can be divided into two large homologous domains, L1 (amino acids 1-155) and L2 (amino acids 313468), separated by a cysteine rich region (amino acids 156-312) (Ward et al., 1995, Prot. Struct. Funct. Genet. 22:141-153). Many determinants of insulin binding seem to reside in the α-subunit. A unique feature of IR is that it is dimeric in the absence of ligand.
The sequence of IR is highly homologous to the sequence of IGF-1R. The sequence identity level varies from about 40% to 70%, depending on the position within the α-subunit. The three-dimensional structures of both receptors may therefore be similar. The crystal structure of the first three domains of IGF-1R has been determined (Garrett et al, 1998, Nature 394:395-399). The L domains consist of a single-stranded right-handed β-helix (a helical arrangement of β-strands), while the cysteine-rich region is composed of eight disulfide-bonded modules.
The β-subunit of the insulin receptor has 620 amino acid residues and three domains: extracellular, transmembrane, and cytosolic. The extracellular domain is linked by disulfide bridges to the α-subunit. The cytosolic domain includes the tyrosine kinase domain, the three-dimensional structure of which has been solved (Hubbard et al., 1994, Nature 372:746-754).
To aid in drug discovery efforts, a soluble form of a membrane-bound receptor was constructed by replacing the transmembrane domain and the intracellular domain of IR with constant domains from immunoglobulin Fc or λ subunits (Bass et at., 1996, J. Biol. Chem. 271:19367-19375). The recombinant gene was expressed in human embryonic kidney 293 cells. The expressed protein was a fully processed heterotetramer and the ability to bind insulin was similar to that of the full-length holoreceptor.
IGF-1 and insulin competitively cross-react with IGF-R and IR. (L. Schäffer, 1994, Eur. J. Biochem. 221:1127-1132). Despite 45% overall amino acid identity, insulin and IGF-1 bind only weakly to each other's receptor. The affinity of each peptide for the non-cognate receptor is about 3 orders of magnitude lower than that for the cognate receptor (Mynarcik, et al., 1997, J. Biol. Chem. 272:18650-18655). The differences in binding affinities may be partly explained by the differences in amino acids and unique domains which contribute to unique tertiary structures of ligands (Blakesley et al., 1996, Cytokine Growth Factor Rev. 7(2):153-9).
Both insulin and IGF-1 are expressed as precursor proteins comprising, among other regions, contiguous A, B, and C peptide regions, with the C peptide being an intervening peptide connecting the A and B peptides. A mature insulin molecule is composed of the A and B chains connected by disulfide bonds, where the connecting C peptide has been removed during post-translational processing. IGF-1 retains its smaller C-peptide as well as a small D extension at the C-terminal end of the A chain, making the mature IGF-1 slightly larger than insulin (Blakesley, 1996). The C region of human IGF-1 appears to be required for high affinity binding to IGF-R (Pietrzkowski et al., 1992, Cancer Res. 52(23):6447-51). Specifically, tyrosine 31 located within this region appears to be essential for high affinity binding. Furthermore, deletion of the D domain of IGF-1 increased the affinity of the mutant IGF-1 for binding to the IR, while decreasing its affinity for the IGF-1R (Pietrzkowski et al., 1992). A further distinction between the two hormones is that, unlike insulin, IGF-1 has very weak self-association and does not hexamerize (De Meyts, 1994).
The α-subunits, which contain the ligand binding region of IR and IGF-1R, demonstrate between 47-67% overall amino acid identity. Three general domains have been reported for both receptors from sequence analysis of the α subunits, L1-Cys-rich-L2. The cysteine residues in the C-rich region are highly conserved between the two receptors; however, the cysteine-rich domains have only 48% overall amino acid identity.
Despite the similarities observed between these two receptors, the role of the domains in specific ligand binding are distinct. Through chimeric receptor studies, (domain swapping of the IR and IGF-R α-subunits), researchers have reported that the sites of interaction of the ligands with their specific receptors differ (T. Kjeldsen et al., 1991, Proc. Natl. Acad. Sci. USA 88:44044408; A. S. Andersen et al., 1992, J. Biol. Chem. 267:13681-13686). For example, the cysteine-rich domain of the IGF-1R was determined to be essential for high-affinity IGF binding, but not insulin binding. When amino acids 191-290 of IGF-1R region was introduced into the corresponding region of the IR (amino acids 198-300), the modified IR bound both IGF-1 and insulin with high affinity. Conversely, when the corresponding region of the IR was introduced into the IGF-1R, the modified IGF-R bound to IR but not IGF-1.
A further distinction between the binding regions of the IR and IGF-1R is their differing dependence on the N-terminal and C-terminal regions. Both the N-terminal and C-terminal regions (located within the putative L1 and L2 domains) of the IR are important for high-affinity insulin binding but appear to have little effect on IGF-1 binding for either IR or IGF-1R. Replacing residues in the N-terminus of IGF-1R (amino acids 1-62) with the corresponding residues of IR (amino acids 1-68) confers insulin-binding ability on IGF-1R. Within this region, residues Phe-39, Arg41 and Pro-42 are reported as major contributors to the interaction with insulin (Williams et al., 1995). When these residues are introduced into the equivalent site of IGF-1R, the affinity for insulin is markedly increased, whereas, substitution of these residues by alanine in IR results in markedly decreased insulin affinity. Similarly, the region between amino acids 704-717 of the C-terminus of IR has been shown to play a major role in insulin specificity. Substitution of these residues with alanine also disrupts insulin binding (Mynarcik et al., 1996, J. Biol. Chem. 271(5):243942; C. Kristensen et al, 1'999, J. Biol. Chem. 274(52):37351-37356).
Further studies of alanine scanning of the receptors suggest that insulin and IGF-1 may use some common contacts to bind to IGF-1R but that those contacts differ from those that insulin utilizes to bind to IR (Mynarcik et al., 1997). Hence, the data in the literature has led one commentator to state that even though “the binding interfaces for insulin and IGF-1 on their respective receptors may be homologous within this interface the side chains which make actual contact and determine specificity may be quite different between the two ligand-receptor systems” (De Meyts, 1994).
Based on data for binding of insulin and insulin analogs to various insulin receptor constructs, a binding model has been proposed. This model shows insulin receptor with two insulin binding sites that are positioned on two different surfaces of the receptor molecule, such that each alpha-subunit is involved in insulin binding. In this way, activation of the insulin receptor is believed to involve cross-connection of the alpha-subunits by insulin. A similar mechanism may operate for IGF-1R, but one of the receptor binding interactions appears to be different (Schäffer, 1994, Eur. J. Biochem. 221:1127-1132).
The identification of molecular structures having a high degree of specificity for one or the other receptor is important to developing efficacious and safe therapeutics. For example, a molecule developed as an insulin agonist should have little or no IGF-1 activity in order to avoid the mitogenic activity of IGF-1 and a potential for facilitating neoplastic growth. It is therefore important to determine whether insulin and IGF-1 share common three-dimensional structures but which have sufficient differences to confer selectivity for their respective receptors. Similarly, it would be desirable to identify other molecular structures that mimic the active binding regions of insulin and/or IGF-1 and which impart selective agonist or antagonist activity.
Although certain proteins are important drugs, their use as therapeutics presents several difficult problems, including the high cost of production and formulation, administration usually via injection and limited stability in the bloodstream. Therefore, replacing proteins, including insulin or IGF-1, with small molecular weight drugs has received much attention. However, to date, none of these efforts has resulted in finding an effective drug replacement.
Peptides mimicking functions of protein hormones have been previously reported. Yanofsky et al., (1996, Proc. Natl. Acad. Sci. USA 93:7381-7386) reported the isolation of a monomer antagonistic to IL-1 with nanomolar affinity for the IL-1 receptor. This effort required construction and use of many phage displayed peptide libraries and sophisticated phage-panning procedures.
Wrighton et al. (1996, Science 273:458464) and Livnah et al. (1996, Science 273:464471) reported dimer peptides that bind to the erythropoietin (EPO) receptor with full agonistic activity in vivo. These peptides are cyclical and have intra-peptide disulfide bonds; like the IL-1 receptor antagonist, they show no significant sequence identity to the natural ligand. Importantly, X-ray crystallography revealed that it was the spontaneous formation of non-covalent peptide homodimer peptides that enabled the dimerization two EPO receptors.
WO 96/04557 reported the identification of peptides and antibodies that bound to active sites of biological targets, which were subsequently used in competition assays to identify small molecules that acted as agonist or antagonists at the biological targets. Renchler et al. (1994, Proc. Natl. Acad. Sci. USA 91:3623-3627) reported synthetic peptide ligands of the antigen binding receptor that induced programmed cell death in human B-cell lymphoma.
Most recently, Cwirla et al. (1997, Science 276:1696-1698) reported the identification of two families of peptides that bound to the human thrombopoietin (TPO) receptor and were competed by the binding of the natural TPO ligand. The peptide with the highest affinity, when dimerized by chemical means proved to be as potent an in vivo agonist as TPO, the natural ligand.